Abstract:The data-driven approach currently dominates the field of Automatic Fault Detection and Diagnosis (AFDD) in HVAC systems. However, a significant concern lies in the prevalent use of labeled experimental and simulation data, which often does not represent real-world operational conditions. This study unveils a comprehensive framework for AFDD in light commercial buildings, effectively leveraging unlabeled raw data extracted directly from their Building BAS. Its main goal is to provide a versatile methodology tailored for real-world applicability. Buildings classified as "light commercial" typically have less than 2,500 square meters of floor area and no more than six stories, such as small offices, medical facilities, banks, small manufacturing facilities, etc. A common feature of these buildings is the fact that the HVAC systems tend to be relatively simple and have similar configurations, thus making it easy to develop scalable and reproducible fault detection methods. The study focuses on a practical case study in a light commercial building HVAC system situated in Montreal, Canada, encompassing a single Air Handling Unit (AHU) and four Variable Air Volume (VAV) reheating boxes to evaluate the framework. This comprehensive framework encompasses a sequence of sub-objectives: creating a sizable, synchronized raw dataset from diverse BAS sensor tags, comprehensive data cleansing to address inconsistencies, developing an anomaly detection method, investigating these anomalies to extract underlying rules, and finally, dataset labeling. An AFDD classification model is then applied to evaluate its ability to distinguish normal from faulty conditions across various fault types. The study highlights the potential of dimensional reduction techniques and unsupervised clustering for effective anomaly detection in light commercial buildings, as well as the power of the Decision Tree classifier for uncovering hidden patterns, especially in anomaly conditions. It also highlights the significance of addressing imbalanced datasets in AFDD and the complexities of detecting sizing-related faults. Despite these challenges, the framework exhibits robust performance in detecting and diagnosing a range of HVAC faults. It offers a systematic and adaptable approach for handling real-world operational data in light commercial building HVAC systems, extendible to other building types, bridging the gap between data-driven methods and practical applications.

A machine learning classifier for automated fault detection and diagnosis (AFDD) of rooftop units, addressing practical challenges of application

Model-based Fault Detection and Diagnosis for HVAC.

Fault detection and diagnosis in light commercial buildings' HVAC systems: A comprehensive framework, application, and performance evaluation

A Machine Learning Framework for Detection and Severity Estimation of Faults for Chillers and Air Handling Units in HVAC systems

Machine Learning-Based Automated Fault Detection and Diagnostics in Building Systems

An Automated Machine Learning Approach for Real-Time Fault Detection and Diagnosis

Transfer Learning for HVAC System Fault Detection

Fault Diagnosis Method Based on Scaling Law for On-line Refrigerant Leak Detection

An automatic and efficient fault diagnosis strategy for air conditioning units by combining attention mechanisms

A data-driven fault detection and diagnosis scheme for air handling units in building HVAC systems considering undefined states

Leveraging Auxiliary Task Relevance for Enhanced Bearing Fault Diagnosis through Curriculum Meta-learning

Development of Anomaly Detectors for HVAC Systems Using Machine Learning

Development of an FDD model for an existing building using transfer learning

How to improve the application potential of deep learning model in HVAC fault diagnosis: Based on pruning and interpretable deep learning method

A Fault Detection Model for Air Handling Units Based on the Machine Learning Algorithms

Research on Fault Diagnosis Strategy of Air-Conditioning Systems Based on DPCA and Machine Learning

A statistical-based online cross-system fault detection method for building chillers

A Hybrid Data-Driven Simultaneous Fault Diagnosis Model for Air Handling Units

Domain-specific Large Language Models for Fault Diagnosis of Heating, Ventilation, and Air Conditioning Systems by Labeled-Data-supervised Fine-Tuning

Advancing Fault Detection in HVAC Systems: Unifying Gramian Angular Field and 2D Deep Convolutional Neural Networks for Enhanced Performance

SRTFD: Scalable Real-Time Fault Diagnosis through Online Continual Learning